def header(header_cells):
"""
Returns a data object after setting it up using the
header.
Parameters
----------
header_cells : list of strings
List of all the column names
Returns
-------
Data object
"""
data = Data()
for col_index, col_name in header_cells.items():
if not re.match(r'\?',col_name):
c = data.use.len
data.use[c] = col_index
data.name[c] = col_name
if re.match(r'[<>\$]', col_name):
data.nums[c] = Num()
else:
data.syms[c] = Sym()
if re.match(r'<', col_name):
data.w[c] = -1
elif re.match(r'>', col_name):
data.w[c] = 1
elif re.match(r'!', col_name):
data.class_col = c
return data
def header(self, cells):
indeps = []
for c0, x in enumerate(cells):
if not re.match('\?', x):
c = len(self._use)
self._use.append(c0)
self.name.append(x)
if re.match(r'[<>\$]', x):
self.nums[c] = Num()
else:
self.syms[c] = Sym()
if re.match(r'<', x):
self.w[c] = -1
elif re.match(r'>', x):
self.w[c] = 1
elif re.match(r'!', x):
self.label_class = c
else:
self.indeps.append(c)
return
def header(self, cells):
for i, x in enumerate(cells):
if not re.match('\?', x):
# print ("printing x", x)
c = len(self.use)
# print ("Printing C", c)
self.use[c] = i
self.name[c] = x
if re.match("[<>$]", x):
self.nums[c] = Num(0)
else:
self.syms[c] = Sym()
#----why are setting goals for length of us? Shouldn't it be for each column?
if re.match("<", x):
self.w[c] = -1
elif re.match(">", x):
self.w[c] = 1
elif re.match("!", x):
self.dclass = c
else:
self.indeps.append(c)
def argmin(c, lo, hi):
cut = None
if (hi - lo > 2 * enough):
l, r = Num(0), Num(0)
for i in range(lo, hi + 1):
r.numInc(rows[i][c])
best = r.sd
for i in range(lo, hi + 1):
x = rows[i][c]
l.numInc(x)
r.numDec(x)
if l.n >= enough and r.n >= enough:
tmp = Num.numXpect(l, r) * 1.04
if tmp < best:
cut, best = i, tmp
return cut
def argmin(self, c, lo, hi):
cut = False
if hi - lo > 2 * self.enough:
l, r = Num(), Num()
for i in range(lo, hi + 1):
r.numInc(self.rows[i][c])
best = r.sd
for i in range(lo, hi + 1):
x = self.rows[i][c]
l.numInc(x)
r.numDec(x)
if l.n >= self.enough and r.n >= self.enough:
tmp = Num.numXpect(l, r) * 1.05 # magic constant
if tmp < best:
cut, best = i, tmp
return cut
def argmin(c, lo, hi):
cut = None
if hi - lo > 2 * enough:
l = Num() # left split
r = Num() # right split
# push everything in the right
for i in range(lo, hi):
r.numInc(rows[i][c])
best = r.sd # currently all data is in right so best is sd on right
# print(best)
# push to the left one by one and keep track of best
for i in range(lo, hi):
x = rows[i][c]
l.numInc(x)
r.numDec(x)
if l.n >= enough and r.n >= enough:
tmp = Num.numXpect(l, r) * 1.05
# print(tmp, x)
if tmp < best:
cut, best = i, tmp
# print(tmp, best)
return cut
def argmin(c, lo, hi):
cut = None
xl, yl = Num(), Num(
) # left split for both features and label cols
xr, yr = Num(), Num(
) # right split for both features and label cols
# push everything in the right
for i in range(lo, hi):
xr.numInc(rows[i][c])
yr.numInc(rows[i][goal])
best_x = xr.sd # currently all data is in right so best is sd on right
best_y = yr.sd # currently all data is in right so best is sd on right
mu = yr.mu
# print(best)
# push to the left one by one and keep track of best
if hi - lo > 2 * enough:
for i in range(lo, hi):
x = rows[i][c]
y = rows[i][goal]
xl.numInc(x)
yl.numInc(y)
xr.numDec(x)
yr.numDec(y)
if xl.n >= enough and xr.n >= enough:
tmp_x = xl.numXpect(xr) * 1.05
tmp_y = yl.numXpect(yr) * 1.05
# print(tmp, x)
try:
if tmp_x < best_x:
if tmp_y < best_y:
cut, best_x, best_y = i, tmp_x, tmp_y
# print(tmp_x, tmp_y, best_x, best_y)
except:
print(tmp_x, tmp_y)
return cut, mu
def computeExpectedValue(self, lo, hi, n, dom_index):
x = Num()
x.bulkAdd(self.rows[lo:hi + 1][dom_index])
return x.sd * x.n / n
def argmin(self, c, lo, hi):
cut = False
xl, xr = Num(), Num()
yl, yr = Num(), Num()
for i in range(lo, hi + 1):
xr.numInc(self.rows[i][c])
yr.numInc(self.rows[i][self.goal])
bestx = xr.sd
besty = yr.sd
mu = yr.mu
if hi - lo > 2 * self.enough:
for i in range(lo, hi + 1):
x = self.rows[i][c]
y = self.rows[i][self.goal]
xl.numInc(x)
xr.numDec(x)
yl.numInc(y)
yr.numDec(y)
if xl.n >= self.enough and xr.n >= self.enough:
tmpx = Num.numXpect(xl, xr) * 1.05 # magic constant
tmpy = Num.numXpect(yl, yr) * 1.05
# print('i=',i,' c=',c,' tmpx=',tmpx,' bestx',bestx)
if tmpx < bestx:
if tmpy < besty:
cut, bestx, besty = i, tmpx, tmpy
return cut, mu